[PATCH] eal: fix cleanup on Windows

[PATCH] eal: fix cleanup on Windows

[Thomas Monjalon]

The memory allocated with _aligned_malloc()
must be released with _aligned_free() on Windows.

The POSIX free() was called in eal_lcore_var_cleanup(),
called in rte_eal_cleanup(), and triggered a heap corruption:
exit status 3221226356 or signal 3221226228 SIGinvalid
with MALLOC_PERTURB_=86

Fixes: 5bce9bed67ad ("eal: add static per-lcore memory allocation facility")

Reported-by: David Marchand <david.marchand@redhat.com>
Signed-off-by: Thomas Monjalon <thomas@monjalon.net>
---
 lib/eal/common/eal_common_lcore_var.c | 5 ++++-
 1 file changed, 4 insertions(+), 1 deletion(-)

diff --git a/lib/eal/common/eal_common_lcore_var.c b/lib/eal/common/eal_common_lcore_var.c
index 0e9e8e4804..a1b2458839 100644
--- a/lib/eal/common/eal_common_lcore_var.c
+++ b/lib/eal/common/eal_common_lcore_var.c
@@ -54,7 +54,6 @@ lcore_var_alloc(size_t size, size_t align)
 		current_buffer = _aligned_malloc(alloc_size, RTE_CACHE_LINE_SIZE);
 #else
 		current_buffer = aligned_alloc(RTE_CACHE_LINE_SIZE, alloc_size);
-
 #endif
 		RTE_VERIFY(current_buffer != NULL);
 
@@ -108,7 +107,11 @@ eal_lcore_var_cleanup(void)
 	while (current_buffer != NULL) {
 		struct lcore_var_buffer *prev = current_buffer->prev;
 
+#ifdef RTE_EXEC_ENV_WINDOWS
+		_aligned_free(current_buffer);
+#else
 		free(current_buffer);
+#endif
 
 		current_buffer = prev;
 	}
-- 
2.46.0

Re: [PATCH] eal: fix cleanup on Windows

[David Marchand]

On Fri, Nov 8, 2024 at 2:08 PM Thomas Monjalon <thomas@monjalon.net> wrote:
>
> The memory allocated with _aligned_malloc()
> must be released with _aligned_free() on Windows.
>
> The POSIX free() was called in eal_lcore_var_cleanup(),
> called in rte_eal_cleanup(), and triggered a heap corruption:
> exit status 3221226356 or signal 3221226228 SIGinvalid
> with MALLOC_PERTURB_=86
>
> Fixes: 5bce9bed67ad ("eal: add static per-lcore memory allocation facility")
>
> Reported-by: David Marchand <david.marchand@redhat.com>
> Signed-off-by: Thomas Monjalon <thomas@monjalon.net>

Reviewed-by: David Marchand <david.marchand@redhat.com>


-- 
David Marchand

RE: [PATCH] eal: fix cleanup on Windows

[Morten Brørup]

Reviewed-by: Morten Brørup <mb@smartsharesystems.com>

Re: [PATCH] eal: fix cleanup on Windows

[Dmitry Kozlyuk]

2024-11-08 14:08 (UTC+0100), Thomas Monjalon:
> The memory allocated with _aligned_malloc()
> must be released with _aligned_free() on Windows.
> 
> The POSIX free() was called in eal_lcore_var_cleanup(),
> called in rte_eal_cleanup(), and triggered a heap corruption:
> exit status 3221226356 or signal 3221226228 SIGinvalid
> with MALLOC_PERTURB_=86
> 
> Fixes: 5bce9bed67ad ("eal: add static per-lcore memory allocation facility")
> 
> Reported-by: David Marchand <david.marchand@redhat.com>
> Signed-off-by: Thomas Monjalon <thomas@monjalon.net>

Acked-by: Dmitry Kozlyuk <dmitry.kozliuk@gmail.com>

Re: [PATCH] eal: fix cleanup on Windows

[Mattias Rönnblom]

On 2024-11-08 14:08, Thomas Monjalon wrote:
> The memory allocated with _aligned_malloc()
> must be released with _aligned_free() on Windows.
> 
> The POSIX free() was called in eal_lcore_var_cleanup(),

Referring to free() as a part of POSIX is true, but a bit misleading, 
since it's also standard C.

> called in rte_eal_cleanup(), and triggered a heap corruption:
> exit status 3221226356 or signal 3221226228 SIGinvalid
> with MALLOC_PERTURB_=86
> 
> Fixes: 5bce9bed67ad ("eal: add static per-lcore memory allocation facility")
> 
> Reported-by: David Marchand <david.marchand@redhat.com>
> Signed-off-by: Thomas Monjalon <thomas@monjalon.net>
> ---
>   lib/eal/common/eal_common_lcore_var.c | 5 ++++-
>   1 file changed, 4 insertions(+), 1 deletion(-)
> 
> diff --git a/lib/eal/common/eal_common_lcore_var.c b/lib/eal/common/eal_common_lcore_var.c
> index 0e9e8e4804..a1b2458839 100644
> --- a/lib/eal/common/eal_common_lcore_var.c
> +++ b/lib/eal/common/eal_common_lcore_var.c
> @@ -54,7 +54,6 @@ lcore_var_alloc(size_t size, size_t align)
>   		current_buffer = _aligned_malloc(alloc_size, RTE_CACHE_LINE_SIZE);
>   #else
>   		current_buffer = aligned_alloc(RTE_CACHE_LINE_SIZE, alloc_size);
> -
>   #endif
>   		RTE_VERIFY(current_buffer != NULL);
>   
> @@ -108,7 +107,11 @@ eal_lcore_var_cleanup(void)
>   	while (current_buffer != NULL) {
>   		struct lcore_var_buffer *prev = current_buffer->prev;
>   
> +#ifdef RTE_EXEC_ENV_WINDOWS
> +		_aligned_free(current_buffer);
> +#else
>   		free(current_buffer);
> +#endif
>   
>   		current_buffer = prev;
>   	}

Reviewed-by: Mattias Rönnblom <mattias.ronnblom@ericsson.com>

Re: [PATCH] eal: fix cleanup on Windows

[Stephen Hemminger]

On Fri,  8 Nov 2024 14:08:21 +0100
Thomas Monjalon <thomas@monjalon.net> wrote:

> The memory allocated with _aligned_malloc()
> must be released with _aligned_free() on Windows.
> 
> The POSIX free() was called in eal_lcore_var_cleanup(),
> called in rte_eal_cleanup(), and triggered a heap corruption:
> exit status 3221226356 or signal 3221226228 SIGinvalid
> with MALLOC_PERTURB_=86
> 
> Fixes: 5bce9bed67ad ("eal: add static per-lcore memory allocation facility")
> 
> Reported-by: David Marchand <david.marchand@redhat.com>
> Signed-off-by: Thomas Monjalon <thomas@monjalon.net>

Is there anyway to use the function attributes to catch this at compile
time (for Gcc builds).

RE: [PATCH] eal: fix cleanup on Windows

[Morten Brørup]

> From: Stephen Hemminger [mailto:stephen@networkplumber.org]
> Sent: Friday, 8 November 2024 17.01
> 
> On Fri,  8 Nov 2024 14:08:21 +0100
> Thomas Monjalon <thomas@monjalon.net> wrote:
> 
> > The memory allocated with _aligned_malloc()
> > must be released with _aligned_free() on Windows.
> >
> > The POSIX free() was called in eal_lcore_var_cleanup(),
> > called in rte_eal_cleanup(), and triggered a heap corruption:
> > exit status 3221226356 or signal 3221226228 SIGinvalid
> > with MALLOC_PERTURB_=86
> >
> > Fixes: 5bce9bed67ad ("eal: add static per-lcore memory allocation
> facility")
> >
> > Reported-by: David Marchand <david.marchand@redhat.com>
> > Signed-off-by: Thomas Monjalon <thomas@monjalon.net>
> 
> Is there anyway to use the function attributes to catch this at compile
> time (for Gcc builds).

I took a look at it, and it boils down to building for a Windows environment, where RTE_EXEC_ENV_WINDOWS is defined, and then the alloc function attributes are disabled.